14 December 2000

Nature 408, 842 - 844 (2000); doi:10.1038/35048548

SONG-YOU HONG*† AND EUGENIA KALNAY‡

* National Centers for Environmental Prediction, Camps Springs, Maryland 20746, USA
† Department of Atmospheric Sciences, Yonsei University, Shinchon, 134 Seoul, Korea
‡ Department of Meteorology, University of Maryland, College Park, Maryland 20742, USA

Correspondence and requests for materials should be addressed to E.K. (e-mail: ekalnay@atmos.umd.edu).

The drought that affected the US states of Oklahoma and Texas in the summer of 1998 was strong and persistent, with soil moisture reaching levels comparable to those of the 1930s 'dust bowl'^{1, 2}. Although other effects of the record-strength 1997–98 El Niño were successfully predicted over much of the United States, the Oklahoma–Texas drought was not³. Whereas the response of the tropical atmosphere to strong anomalies in sea surface temperature is quite predictable, the response of the extratropical atmosphere is more variable^{4, 5}. Here we present results from mechanistic experiments to clarify the origin and maintenance of this extratropical climate extreme. In addition to global atmospheric models^6-11, we use a regional model^{12, 13} to isolate regional climate feedbacks. We conclude that during April and May 1998, sea surface temperature anomalies combined with a favourable atmospheric circulation to establish the drought. In June–August, the regional positive feedback associated with lower evaporation and precipitation contributed substantially to the maintenance of the drought. The drought ended in the autumn, when stronger large-scale weather systems were able to penetrate the region and overwhelm the soil-moisture feedback. Our results show the potential for numerical models including appropriate physical processes to make skilful predictions of regional climate.

The strong Oklahoma–Texas drought had already been established by May 1998, with high temperatures and a low-precipitation weather pattern typical of July and August. It persisted until early October throughout most of Oklahoma and Texas, except in southern Texas, where it rained in August and September. The drought was not predicted by the US operational seasonal forecasts made in March 1998³ at the National Centers for Environmental Prediction (NCEP). It is important to establish the mechanisms that originated, maintained and ended the drought in order to establish whether such extreme drought is predictable.

There are three main mechanisms that can produce extended atmospheric anomalies: (1) sea surface temperature (SST) anomalies, (2) soil-moisture anomalies, and (3) atmospheric initial conditions favourable to such a climate extreme even in the absence of surface forcing (that is, internal forcing). Tropical SST anomalies, such as those associated with El Niño and La Niña episodes, have a profound effect in tropical and subtropical regions⁴. A drought can also be enhanced through a regional positive feedback: dry soil reduces local evaporation and therefore can reduce precipitation^{6, 7, 10, 11}. It is also possible that the atmosphere in the late spring of 1998 was already 'primed' to produce a drought in the Oklahoma–Texas region even without anomalies of SST or soil moisture⁸.

It would seem natural to "blame El Niño" of 1997/1998 for the drought, but there are problems with this simple assumption. First, warm El Niño/Southern Oscillation (ENSO) episodes tend to be correlated with high precipitation in April–June over Oklahoma and Texas^14-16. In particular, the drought of 1988 in the Great Plains has been attributed to La Niña (cold) conditions^{6, 9, 17}. Furthermore, the warm ENSO episode faded in June 1998, and was quickly replaced with strong La Niña (cold SST) conditions in the equatorial Pacific Ocean. Conversely, soil-moisture anomalies alone could not have triggered the drought, because the Oklahoma–Texas region had above-normal precipitation during the early spring, so that the soil was wetter than normal in the spring of 1998.

For this reason, we tested alternative hypotheses, using numerical models and mechanistic experiments. The hypotheses were: (1) both atmospheric conditions in the spring and the SST anomalies associated with the waning warm ENSO episode were crucial in the establishment of the anomalous Oklahoma–Texas drought conditions during the spring of 1998. (2) The consequent deficits in the soil moisture contributed to the maintenance of the drought throughout the summer and early autumn through a local positive feedback between the atmosphere and the land. (3) The drought ended in the autumn, when stronger weather systems were able to penetrate this region and overwhelm the soil-moisture feedback.

Tropical SST anomalies have a global-scale effect, whereas soil-moisture anomalies have a much more regional effect. Because the two mechanisms invoked for the drought during the spring and summer are quite different, we need different modelling approaches to test the first two hypotheses. To test the first hypothesis we integrated a version of the NCEP global spectral model (GSM)¹⁸ with climatological and observed SSTs, and with 1998 and 1993 initial conditions. For the second and third hypotheses, based on regional feedbacks, we needed to maintain the observed global circulation, so we opted for running the NCEP regional spectral model (RSM)^{12, 13}. The horizontal resolution of the GSM and RSM is about 200 km and 50 km, respectively. We first checked whether the NCEP/NCAR 50-year reanalysis¹⁹ anomalies in precipitation were reasonable, as they determine the soil-moisture anomalies. Figure 1 compares the observed and the reanalysis precipitation centred over the Oklahoma–Texas region (which we define as the region bounded by 29° N–37° N and 105° W–91° W) from mid-March to mid-September 1998. It shows that the reanalysis 6-hour forecast reproduced very well the amount and timing of the observed precipitation. Both this, and other comparisons with surface and soil observations from the Oklahoma Mesonet¹ of surface observations, provide confidence that the reanalysis can indeed be used for this purpose.

Figure 1 Comparison of observed precipitation and that predicted by the NCEP/NCAR reanalysis.   Full legend   High resolution image and legend (37k)

To check the influence of the SST anomalies, we performed global ensembles of five forecasts starting from initial conditions corresponding to the 5 days of 26–30 April 1998, and verified during the month of May 1998. Two experiments were run: OSST, with observed SST in 1998; CSST, with climatological SST. To test the effect of the initial conditions an additional ensemble experiment denoted '1993' was run with SST and soil moisture anomalies for 1998, as in OSST, but with initial conditions corresponding to 1993 (a flood year).

Figure 2a shows the observed deficit in precipitation for May 1998, and Fig. 2c shows anomalous low-level moisture transport from the Gulf of Mexico into the Oklahoma–Texas region. They suggest that the drought in May was associated with a lower than normal input of moisture into the region from the low-level jet coming from the Gulf^{20, 21}. Figure 2b and d show the effect of the SST anomalies present during May. They indicate that the SST anomalies contributed to the establishment of the drought by producing a significant reduction in precipitation, enough to explain about 60% of the deficit observed at that time. Similar experiments were performed for April and June 1998 and the results, summarized in Table 1, generally confirm those obtained for May 1998.

Figure 2 Comparison of the model impact of SST anomalies (OSST - CSST) with the observed anomalies for May 1998.   Full legend   High resolution image and legend (209k)

The extent to which the atmospheric conditions in 1998 were already 'primed' to generate the drought is shown in Fig. 3. It is clear that in 1998 the atmospheric circulation would have resulted in lower precipitation than in 1993, even if the surface forcing from SST and soil moisture were the same in both years. The reduction is about 40% of the observed deficit (Table 1).

Figure 3 Effect of the initial conditions on the drought.   Full legend   High resolution image and legend (87k)

We tested the second hypothesis (maintenance of the drought through soil-moisture positive feedback) using the regional spectral model. Reanalysis boundary conditions and SSTs were used to drive the RSM during April–September 1998. Only the bottom (deep) soil-moisture model layer was updated from the reanalysis every 24 hours during the model integration period, while the soil moisture in the top (shallow) layer was predicted, to maintain balance with the atmospheric forcing.

Three types of runs were made with the regional spectral model. The first two, CONTROL and CLIMAT, had deep soil moisture replaced by the reanalysis soil moisture for 1998 and the reanalysis 50-year climatology, respectively. As these RSM experiments changed only the lowest soil model layer, they probably underestimated the positive feedback on the drought. To test the effect of a more intense deficit in the soil moisture, we also performed an experiment (DRY) in which the lower-soil model moisture was maintained at the wilting point. The results are presented in Fig. 4. They suggest that soil-moisture anomalies alone would have in fact resulted in an increase in precipitation during April and May, as the early spring had above-normal precipitation and the soil was wetter than normal at the beginning of this period. However, during the summer, soil-moisture feedback could maintain an extreme drought through local feedback. The fact that by September the precipitation recovers to near CLIMAT levels even in the DRY experiment, supports the third hypothesis—that the end of the drought in the autumn was due to stronger events overwhelming the regional positive feedback.

Figure 4 Effect of the soil-moisture feedback on the drought.   Full legend   High resolution image and legend (32k)

We introduced in this study the use of a regional model nested within the observed large-scale forcing in order to isolate the local feedback mechanism between soil moisture and precipitation. The results suggest that this extratropical climate extreme is not simply rooted in a tropical SST forcing and/or soil-moisture anomalies^{2, 3, 6, 7, 9, 17}. Rather, it is due to a nonlinear coupling of SST anomalies and atmospheric internal forcing organized in the previous months, and a physical interaction with soil-moisture anomalies. At present, statistical seasonal climate predictions are still more skilful than predictions of numerical models²². Our results suggest that research with dynamic models, including appropriate physical processes and improved initial conditions, have the potential to overcome this disadvantage.

Received 0 May 2000;accepted 3 October 2000

References

1.	Basara, J. B., Arndt, D. S., Johnson, H. L., Brotzge, J. G. & Crawford, K. C. An analysis of the drought of 1998 using the Oklahoma Mesonet. Eos 79, 258 (1998).
2.	Bell, G. D. et al. Climate assessment for 1998. Bull. Am. Meteorol. Soc. 80, S1-S48 (1999). | Article | ISI |
3.	Barnston, A. G. et al. NCEP forecasts of the El Niño of 1997-1998 and its U.S. impacts. Bull. Am. Meteorol. Soc. 80, 1829-1852 (1999). | Article | ISI |
4.	Wallace, J. M. et al. On the structure and evolution of ENSO-related climate variability in the tropical Pacific: Lessons from TOGA. J. Geophys. Res. 103, 14241-14259 (1998). | ISI |
5.	Kumar, A. & Hoerling, M. Annual cycle of Pacific-North American predictability associated with different phases of ENSO. J. Clim. 11, 3295-3308 (1998) | Article | ISI |
6.	Atlas, R., Wolfson, N. & Terry, J. The effects of SST and soil moisture anomalies on GLA model simulations of the 1988 US summer drought. J. Clim. 6, 2034-2048 (1993). | Article | ISI |
7.	Wolfson, N., Atlas, R. & Sud, Y. C. Numerical experiments related to the 1980 heat wave. Mon. Weath. Rev. 115, 1345-1357 (1987). | Article | ISI |
8.	Mo, K., Zimmerman, J. R., Kalnay, E. & Kanamitsu, M. A GCM study of the 1988 United States drought. Mon. Weath. Rev. 119, 1512-1532 (1991). | Article | ISI |
9.	Trenberth, K. E., Branstator, G. W. & Arkin, P. A. Origins of the 1988 North American drought. Science 242, 1640-1645 (1988). | ISI |
10.	Shukla, J. & Mintz, Y. The influence of land-surface evaporation on the earth's climate. Science 215, 1498-1500 (1982). | ISI |
11.	Koster, R., Suarez, M. & Heiser, M. Variance and predictability of precipitation at seasonal/to interannual timescales. J. Hydrometeorol. 1, 26-46 (2000). | Article | ISI |
12.	Juang, H.-M. H., Hong, S.-Y. & Kanamitsu, M. The NCEP regional spectral model: An update. Bull. Amer. Meteorol. Soc. 78, 2125-2143 (1997). | Article | ISI |
13.	Roads, J. & Chen, S. Surface water and energy budgets in the NCEP Regional Spectral model. J. Geophys. Res. (in the press).
14.	Ting, M. & Wang, H. Summertime U.S. precipitation variability and its relation to Pacific sea surface temperature. J. Clim. 10, 1853-1873 (1997). | Article | ISI |
15.	Montroy, D. L. Linear relation of central and eastern North American precipitation to tropical Pacific sea surface temperature anomalies. J. Clim. 10, 531-548 (1997). | Article |
16.	Montroy, D. L., Richman, M. B. & Lamb, P. Observed nonlinearities of monthly teleconnections between tropical Pacific sea surface temperature anomalies and central and eastern North American prediction. J. Clim. 11, 1812-1835 (1998). | Article | ISI |
17.	Chen, P. & Newman, M. Rossby wave propagation and the rapid development of upper-level anomalous anticyclones during the 1988 US drought. J. Clim. 11, 2491-2504 (1998). | Article | ISI |
18.	Caplan, P. et al. Changes to the 1995 NCEP operational medium-range forecast model analysis-forecast system. Weath. Forecast. 12, 581-594 (1997) | Article | ISI |
19.	Kalnay, E. et al. The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc. 77, 437-471 (1996) | Article | ISI |
20.	Bonner, W. D. & Paegle, J. Diurnal variations in the boundary-layer winds over the south central United States in summer. Mon. Weath. Rev. 98, 735-744 (1970) | ISI |
21.	Schubert, S. D., Helfand, H. M., Wu, C.-Y. & Min, W. Subseasonal variations in warm-season moisture transport and precipitation over the central and eastern United States. J. Clim. 11, 2530-2555 (1998). | Article | ISI |
22.	Anderson, J. et al. Present-day capabilities of numerical and statistical models for atmospheric extratropical seasonal simulation and prediction. Bull. Am. Meteorol. Soc. 80, 1349-1361 (1999). | Article | ISI |

Acknowledgements. We thank M. Richman, K. Crawford and J. Basara for helpful discussions.